Contract Portfolio Intelligence

How Do You Find Out What Your
Contracts Actually Say?

What extraction returns from an executed portfolio, why contract wording cannot be compressed, and where the signed page takes over.

Built from real buyer questions in our sales meetings

An agreement you cannot query is an obligation you accepted and forgot. It runs on schedule whether or not anybody reads it. Executives rarely tell us their contracts are lost; they tell us something sharper — nobody in the building can answer a question that spans them.

Direct Answer

Ingest the executed agreements as a governed corpus, then query them by obligation rather than by filename. The questions that matter run across the portfolio, not down a single file: which agreements grant an audit right, which expire inside ninety days, which predate the regulation you are now measured against. Iternal built its contract intelligence pipeline for that shape of question — extraction pulls terms, parties, dates, obligations and compliance flags out of every document, and a second pass analyzes the structured set as one body of evidence.

The limit: the step that makes contracts queryable can move a word. Iternal reports that this form of ingestion may change some of the semantic words while the underlying message is preserved. For a portfolio question — which agreements carry this obligation — that is tolerable, because the output is a ranked list of documents to open. For the wording of a clause you are about to rely on it is not. The answer routes you to the executed document, and the executed document is what you quote.

Contracts cannot be summarized away, so ask which ingestion path runs on yours. Iternal maintains a Blockify variant built for legal agreements precisely because the exact wording has to survive the trip. Get the field list in writing before you get a price, confirm every structured result links back to the page it came from, and ask where the processing runs — work at portfolio scale is server or private-cloud work, not a laptop job.

Reading a portfolio and drafting a document are different jobs. Querying what your signed agreements already contain is one job; filling in a document you are about to send is another. For more information on filling in your own template, visit the proposals and RFP page; for an outside request, visit the records and evidence page.

Why Contract Text Survives Ingestion Differently

Most enterprise content tolerates compression: a policy deck or a technical manual can lose its phrasing and keep its meaning. An agreement cannot. The operative force lives in the words and in the order they were signed in, which is why Iternal treats legal documents as a class that must maintain its language.

The variant is the point. Iternal ships Blockify variants tuned to different content, and one exists for legal agreements, court records and contracts, where the actual words carry the obligation; that variant keeps the exact precise wording. Blockify also tracks block provenance, so a structured result carries the file location it came from and a reader can open the source and read its raw text — the structured layer is an index into the record, never the record itself.

Three Ways a Portfolio Goes Dark

Buyers describe the same failure arriving by three routes. Each is survivable alone; together they produce an archive that binds the company and informs nobody inside it.

  • Nobody has read it. Historic statements of work and master agreements pile up for years and a lot of them are a mess. Capturing metadata means manual data entry that never gets done. One lawyer reads one agreement, forms a judgment, and the knowledge leaves when the matter closes. Compliance flagging then runs on luck.
  • Renewal outruns review. Agreements renew unchanged because renegotiating takes too long and there is no manpower to review them, and notice of a renewal arrives too late to act on. Older paper signed before the privacy regimes now applied to it stays in force outside the required risk profile. Rebates and payment terms go unclaimed for the same reason: nobody can read that volume.
  • Capacity cannot be bought. No budget funds top legal talent reading every agreement in a large portfolio every year, and cost caps how many reviewers get staffed on a large document set. As one buyer put it, there is more paperwork than there are lawyers.

None of the three is a filing problem. The documents are stored; what is missing is a way to ask them a question.

What Extraction Returns From a Single Agreement

Take one signed agreement and ask the four questions a general counsel asks first. The table pairs each with what Iternal pulls out of the document, and with the decision it changes once the field exists for every agreement:

The question What extraction returns What it changes at portfolio scale
When does it renew? Effective date, expiry date, days remaining, and whether the agreement has lapsed. A live view of what falls due at 90, 60 and 30 days, ranked by risk instead of found by accident.
How long is the notice window? The clause setting the notice period, read against the renewal date it governs. Separates renewals you can still renegotiate from those past the point of choice.
What is the liability position? Limitation of liability, indemnification, audit rights and the reporting duties attached to each. Agreements outside your standard position become ranked exceptions rather than surprises.
Can it be assigned? Whether the agreement is freely assignable and survives a change of control. Marks what needs counterparty consent before a transaction can close.

Two passes, not one. Per-agreement extraction runs first; analysis of the structured set runs second. Iternal compares that set against your own reference for a compliant agreement, marks the gaps and red flags on each document, and produces an executive view of which agreements are clean and which carry exposure. The destination is configurable too: a spreadsheet scoring every agreement against a compliance test is as valid a deliverable as an interface.

A Contract Review Checklist for an Executed Portfolio

A contract review checklist is the fixed set of terms every agreement is read for: parties and signature status, term and renewal dates, notice windows, payment and rebate triggers, liability and indemnity caps, audit and reporting duties, assignment and change of control, governing law, and termination rights.

The checklist is what makes a portfolio comparable. Read nine agreements ad hoc and you get nine opinions; read them against the same nine lines and you get a table you can sort. Iternal configures the extraction fields directly from a list like this one, which is why the field list belongs in writing before a pipeline is built rather than after.

  1. 01
    Parties and signature status

    Legal entity names on both sides, the signing authority used, and whether every counterpart is executed.

  2. 02
    Term, expiry and renewal mechanics

    Effective date, end date, whether it renews automatically, and the length of each renewal period.

  3. 03
    Notice windows

    How many days of notice a non-renewal or termination needs, and the date that window opens.

  4. 04
    Price, payment and rebate triggers

    Payment terms, uplift or indexation clauses, and any volume tier that earns money back.

  5. 05
    Liability and indemnity

    The cap, the carve-outs above it, and which indemnities run in each direction.

  6. 06
    Audit, security and reporting duties

    Audit rights, security commitments, breach-notification clocks and anything owed on a fixed timetable.

  7. 07
    Assignment and change of control

    Whether the agreement transfers freely, and which counterparties must consent before a transaction closes.

  8. 08
    Governing law and dispute route

    Jurisdiction, venue, and whether disputes go to arbitration before they reach a court.

  9. 09
    Termination and exit

    Termination for cause and for convenience, transition assistance, and what happens to data on exit.

Two rules keep the list usable. Record the answer and the location it came from, so every line can be opened on the signed page. And keep the list short enough that it is applied to every agreement rather than to the interesting ones — a checklist nobody finishes produces the same blind portfolio as no checklist at all.

The Contract Review Process Flowchart, Stage by Stage

A contract review process flowchart runs six stages: intake and triage, extraction against the checklist, comparison to your standard position, exception routing to a reviewer, negotiation or acceptance, and execution with the obligation calendar updated. Each stage names an owner, so a document never waits on nobody.

  1. Stage 1
    Intake and triage

    Owner: Requester or contract owner

    Leaves the stage: The document, its counterparty, its value band and the date it is needed by.

  2. Stage 2
    Extraction against the checklist

    Owner: The pipeline

    Leaves the stage: Every checklist field filled, each one carrying the page it was read from.

  3. Stage 3
    Comparison to your standard position

    Owner: The pipeline

    Leaves the stage: Each term marked as inside your position, acceptable, or an exception.

  4. Stage 4
    Exception routing

    Owner: Legal operations

    Leaves the stage: Only the flagged terms reach a reviewer, ranked by the exposure they carry.

  5. Stage 5
    Negotiation or acceptance

    Owner: The named reviewer

    Leaves the stage: A redline proposal with the reasoning attached, or a recorded decision to accept.

  6. Stage 6
    Execution and calendar update

    Owner: Contract owner

    Leaves the stage: The executed file back in the corpus, with its renewal and notice dates live.

Drawn this way, the machine work and the judgment work separate cleanly. Stages two and three are the ones a pipeline does at volume; stages four and five are the ones that need a person and a reason. The failure most teams describe is not a missing stage — it is stage six never happening, so the executed file leaves the process and the renewal date leaves with it.

The Contract Playbook AirgapAI Reviews Against

A contract playbook records three positions for every term you negotiate: the one you prefer, the one you accept without escalation, and the line you will not cross. Load it as a governed corpus and AirgapAI reads each agreement against your positions rather than against internet averages.

What goes in it. One row per term from the checklist above. For each row: your preferred wording, the fallback you accept without escalation, the point that becomes an escalation, and one sentence of reasoning a reviewer can quote back to a counterparty. The reasoning column is the part teams skip and the part that makes the output usable, because a flag without a reason is another thing to look up.

Running it over the executed portfolio. The same playbook that governs the next agreement grades the ones already signed. A sweep reads every executed document against every row and returns three lists: agreements inside your positions, agreements carrying an accepted fallback, and agreements outside the line — the last being the set worth a renegotiation slot at renewal, and where the procurement cost savings in a signed base usually sit. AirgapAI handles the reviewer-facing half on the reviewer's own machine, so privileged paper is read where it already sits; a sweep across the whole archive is server or private-cloud work.

The playbook is the product. Whatever interface you buy, the contract review system you end up with is your positions, written down, applied consistently. Encode a weak fallback and it is defended at scale. Encode nothing and every reviewer re-derives the firm's position from memory, which is the state most portfolios are in before anyone writes the first row.

Checking an Agreement Against Your Standard Position

Reading the portfolio is one job; policing the next agreement is another. Legal teams described reviewers working through customer paper continuously to check that what the company sells stays in compliance, while every regulated counterparty insists on its own bespoke template. The agreement carrying your preferred language, one buyer noted, is the one no client accepts.

Why a general assistant stalls here. A general-purpose model has no access to your fallback positions, so it grades a clause against the average agreement on the internet rather than against yours. It compresses language that has to stay intact, and it reads one document at a time, so it cannot rank a portfolio. Legal teams told us plainly that such tools cannot be relied on for agreements and policies.

What a configured workflow does instead. Iternal configures the review workflow with your most-favorable, medium and willing-to-accept position on each term, recommends the red lines and explains the reasoning behind each; Iternal runs its own master agreements and NDAs through it. Documents stop being read raw, and reviewers read the proposed red lines instead.

The limit is structural. Playbook redlining is only ever as good as the playbook behind it — encode a weak fallback and the machine defends it at scale, and a person still signs. Nor is this device-class work: analysis at portfolio scale runs on a server or a dedicated private-cloud environment, because the on-device ingestion utility exists for small jobs rather than an archive. For more information on where processing may legally sit, visit running AI on data that cannot leave.

Where Contract Review Automation Fits Alongside the Tools You Have

Contract review automation covers three distinct jobs, and most teams need more than one tool. Drafting and redlining happen inside the editor. Extraction across a signed portfolio needs a platform built for volume. Keeping executed agreements governed afterward is a separate system. Match the tool to the job.

The category is strong and the leaders are good at different halves of it. Spellbook sits inside Microsoft Word and is the natural pick for drafting and redlining the agreement in front of you. Harvey is the dedicated legal platform of the group and the strongest fit for reading across a whole signed base. Iternal is complementary to both: AirgapAI is the option for material that cannot leave the building, and it is chosen on where the processing runs rather than on features. For the fuller side-by-side visit the legal AI options comparison; for governing the executed agreements after signature visit the AI for contract management page; and to size the reviewer hours a portfolio actually costs, use the AI Contract Review Calculator.

Four Answers to Get in Writing First

Run the evaluation the way a clause gets drafted. Iternal states that its legal-document ingestion variant preserves the exact precise wording, and that this form of ingestion may move some semantic words while keeping the underlying message. Both describe paths through one pipeline; the path pointed at your archive is the one to pin down.

Pin it down: questions for your evaluation
  • Which ingestion variant runs on our executed agreements, and can we see a clause come out the other side word for word?
    Whether the wording-preserving legal variant is the path applied to your archive, confirmed on your own document.
  • Show us one extracted obligation traced back to the exact page of the signed PDF it came from.
    That the structured layer works as an index into the record, so a lawyer quotes the executed page.
  • Which fields will you extract for us, and what happens when we want to add one after the pipeline is live?
    The configuration boundary, before a new obligation type becomes an unscoped change request.
  • Where does the processing run, which environment holds the documents while it runs, and who has access?
    Residency, isolation and access, while a pilot can still run inside your own tenant on your own material.
Answered elsewhere
FAQ

FAQ: Reading an Executed Contract Portfolio

Ingest the executed documents as a governed corpus and query them by obligation instead of by filename. Extraction pulls parties, dates, governing law, obligations and compliance flags from every agreement, and a second pass analyzes the whole set, so a question such as which agreements grant an audit right returns a ranked list.

Yes, and those fields are why organizations start. Iternal extracts audit rights, rebate eligibility tied to purchasing tiers, breach-notification duties, payment obligations and schedules, and reporting owed on a fixed timetable, attributing each duty to the party carrying it. Extraction targets are configured per organization, so agree the field list in writing before the build.

Make the renewal calendar a property of the portfolio rather than of institutional memory. Once effective dates, expiry dates, days remaining and notice periods are extracted from every agreement, the same pipeline shows what falls due at 90, 60 and 30 days and flags terms that no longer meet current requirements.

Iternal configures the review workflow with your most-favorable, medium and willing-to-accept position on each term, recommends the red lines and explains the reasoning behind each. Documents stop being read raw; reviewers read the proposed red lines instead. The limit is structural: output is only as good as the positions you encoded, and a person still signs.

Stop treating full review as the unit of work. No budget funds senior legal talent reading an entire portfolio every year, and headcount grows more slowly than paper does. Extraction makes triage possible: every agreement is machine-read for the fields that carry risk, and scarce reviewer hours go to the ranked exceptions.

Go back to the signed document. Iternal reports that this form of ingestion may change some of the semantic words even though the underlying message is preserved, so the structured layer is an index into the record rather than the record. Iternal also ships a Blockify variant for legal agreements that keeps the exact precise wording — confirm it runs on your archive, and quote the executed page.

Nine lines, applied to every agreement rather than to the interesting ones: parties and signature status, term and renewal mechanics, notice windows, price and rebate triggers, liability and indemnity, audit and reporting duties, assignment and change of control, governing law, and termination and exit. Record where each answer was found so any line can be opened on the signed page.

Six: intake and triage, extraction against the checklist, comparison to your standard position, exception routing to a named reviewer, negotiation or acceptance, and execution with the renewal calendar updated. Stages two and three run at volume; four and five need a person and a reason. Stage six is the one teams skip, which is how a renewal date leaves the process with the file.

Start With One Question That Spans the Archive

Pick a question you cannot answer today — which agreements let a counterparty audit us, which renew before the quarter closes — and run a pilot on a sample of your own executed documents. Either the ranked list comes back and the line to the signed page holds, or it does not. One question on your own paper settles more than a year of proposals.

John Byron Hanby IV
About the Author

John Byron Hanby IV

CEO & Founder, Iternal Technologies

John Byron Hanby IV is the founder and CEO of Iternal Technologies, a leading AI platform and consulting firm. He is the author of The AI Strategy Blueprint and The AI Partner Blueprint, the definitive playbooks for enterprise AI transformation and channel go-to-market. He advises Fortune 500 executives, federal agencies, and the world's largest systems integrators on AI strategy, governance, and deployment.